Driving Catalytic Innovation: The Development and Application of Machine Learning Force Fields

Catalytic reactions lie at the heart of the modern chemical industry, where their efficiency being determined by microscopic interactions occurring on the surface of the catalyst. A comprehensive understanding of these mechanisms in atomic-scale is imperative for the rational design and process optimization of chemical reactions. However, computational simulations have long grappled with inherent limitations: high-accuracy first-principles methods demand exorbitant computational resources, whereas classical force fields often fail to capture the nuanced dynamics of chemical reactions. The emergence of machine learning force fields (MLFFs) has led to significant advancements in addressing this bottleneck, offering a promising balance between precision and efficiency. This review provides a systematic overview of the evolution of MLFFs methodologies in practical catalytic applications, and highlights their respective advantages as well as the persisting challenges. By consolidating contemporary research trends and unresolved obstacles, this work provides valuable insights for subsequent researchers by synthesizing existing approaches and challenges, contributing to the refinement of MLFFs computational algorithms and their broader application in complex catalytic systems.